Comments on: PostgreSQL 17-18 major upgrade – blue-green migration with minimal downtime https://www.dbi-services.com/blog/postgresql-17-18-major-upgrade-blue-green-migration-with-minimal-downtime/ Tue, 17 Jun 2025 07:05:24 +0000 hourly 1 By: Adrien Obernesser https://www.dbi-services.com/blog/postgresql-17-18-major-upgrade-blue-green-migration-with-minimal-downtime/#comment-4045 Tue, 17 Jun 2025 07:05:24 +0000 https://www.dbi-services.com/blog/?p=38391#comment-4045 Hi, thank you for your comment.
The pg_promote() is part of my usual routine. I left it there to remind people to check that part and for myself. Usually during a migration done manually I put the replica in read-only mode to avoid writes in critical phases from the application or any other users that would have access to the replica since the test began.
Regarding the sequences, yes you would need to, as noted in the blog post pg_createsubscriber is under the same restrictions as logical replication on which it depends. I could have added a reminder though.
There are plenty of already well written procedures/scripts on that.

]]> By: Krisavi https://www.dbi-services.com/blog/postgresql-17-18-major-upgrade-blue-green-migration-with-minimal-downtime/#comment-4043 Mon, 16 Jun 2025 17:40:30 +0000 https://www.dbi-services.com/blog/?p=38391#comment-4043 Odd that the pg18 wasn’t already read/write after pg_createsubscriber and needed promote in the end.

Didn’t you need to sync sequences after: Here I stopped any new writes on the primary and checked the current_lsn. ?

]]>